For many years FOI has conducted research on automatic target detection for different types of sensors. The purpose is to facilitate the work of sensor operators and to increase efficiency and quality in reconnaissance. The field study Image Processing for Combat Vehicles aims at increasing the knowledge in the Swedish Armed Forces about how digital image processing can improve the crew's ability to detect threats early. A large number of scenarios were developed in collaboration with the Armed Forces to enable evaluation of the value added by detection algorithms in combat vehicles. Based on these scenarios sensor data were collected with visual and thermal cameras corresponding to the actual sensors used in combat vehicles in the Swedish Armed Forces. Two detection algorithms were considered in this field study - a motion detection algorithm and a shape detection algorithm. Both show very good potential for threat detection. The motion detection algorithm is able to detect very subtle movements and an evaluation shows that when applied to thermal imagery in an urban scenario it detects a significantly larger number of targets than a combat vehicle crew. The shape-based detection algorithm performs well in rural terrain but currently generates too many false alarms in urban areas. This is primarily due to the algorithm being trained mainly on data acquired in rural terrain; more training data from builtup environments would improve the performance in urban areas. Of the two algorithms the motion detection one currently performs best, and it is believed that the benefits today are greatest with this algorithm. This algorithm works very well in scenarios when the vehicle is standing still, e.g. at a short break or observation post outside an urban terrain. This field study has also shown that automatic motion-based detection of people with the visual cameras on Patgb 360 is efficient up to ranges of about 55 m in good lighting conditions. In order to achieve the best possible performance for the shape-based detection algorithm, e.g. early target detection and fewer false alarms, sensor data should be compressed as little as possible as the algorithms need video signals with high bit resolution; analog signals (or digital signals reproduced from an analog ones) are less suitable for image processing because the conversion has a destructive effect on the original images which cannot be reconstructed through digitization of the signal. Hence, image processing should take place before the signal from the cameras has been converted to an analog signal.